The Lupin Digital Health engagement is covered by a non-disclosure agreement. Enter the passcode to read it — it is on my resume, and I am happy to share it on request via LinkedIn.
Client engagement via Auriga IT, where I have been on payroll since 2015. Lyfe is India's first evidence-based cardiac digital therapeutics platform — FDA and CE approved wearables, medication and vitals tracking, human care managers and coaches, and 24×7 emergency assistance. It launched publicly in January 2023, five weeks after I joined. Lupin's own problem statement: one in three patients show poor monitoring, adherence and awareness after a coronary intervention.
I was hired to do two things: raise adherence in a live product, and extend the platform from post-heart-attack patients to heart failure patients.
The second one carried the hard problem. Heart failure care depends on the 6-minute walk test, a standard clinical assessment of functional capacity. It is normally administered in a hospital corridor with a clinician present, because patients can become breathless, dizzy or symptomatic during it. Our version had to run on a phone, at home, with nobody clinical in the room.
Before designing anything, I asked to sit with one of the cardiologists on the doctor panel to understand the test and where it becomes dangerous. That conversation, not the brief, produced the design.
Lead designer on the patient app, working with a PM, a doctor panel, and a research team that supplied insights.
Direct access to patient research. I sat in on some interviews but never held the recordings or reports, so I cannot claim that evidence as mine. There was no usability testing with cardiac patients in my time there.
That absence is the defining constraint of this project, and it is why the clinician session mattered so much. When I could not go to users, I went to the people who understood the risk.
The rest of this focuses on the walk test, because it is the one where getting it wrong has a clinical cost.
A clinical test moved to an unsupervised setting loses the thing that made it safe, which is not the corridor or the stopwatch. It is the clinician standing there deciding whether this patient should continue.
What replaces the clinician?
Alongside it sat a second tension. A patient who stops early, or posts a worse distance than last month, has produced clinically meaningful information. They also happen to be a frightened person who has recently been very unwell. Reassurance is right for the patient. Suppressing the signal is wrong for their care. The design could not choose one.
Each covering something the missing supervisor would have done.
Before anything else, a single question: are you feeling comfortable to walk? A "no" does not warn or nag. It ends the session with a heart illustration and plain reassurance that the test should only be taken when feeling well, and an invitation to come back later. No guilt, no dark pattern, no path around it.
A clinician screens the patient before the test. This is that screen, self-administered.
The instruction screen requires the patient to take the test with a family member or caretaker present. This is the decision the whole flow rests on. I brought the clinician's account of why the test is supervised back to the team, and we landed on substituting a lay human for a clinical one. Not equivalent, and it does not need to be. A caretaker cannot assess functional capacity, but they can notice someone is in trouble and act, which is the part that actually protects the patient.
It also fits the product rather than fighting it. Lyfe is built on keeping the caretaker connected to the patient's progress, so the person we needed in the room was already part of the model.
Illustrated, one rule per line, minimal text: only take the test when feeling well, stop for chest pain or dizziness, have someone with you, don't talk during the test, wear comfortable clothes and shoes, don't eat or drink, follow the on-screen instruction and keep the phone with you while walking.
The published 6MWT protocol permits a patient to rest and then resume, with distance measured across the full six minutes. Our doctor panel directed that a patient who needs to stop should end the test rather than resume it, for accuracy of the result. That is a clinician-directed deviation, not a design simplification, and the case study says so.
The stop control sits under the timer for the entire six minutes, in the same position throughout. It confirms once, "are you sure you want to stop the test", with the continue option worded as an invitation rather than a warning.
A patient who feels unwell should never have to hunt for the exit or read a scary modal to use it.
A ten-item symptom checklist follows the walk. Each with an icon, because this app serves patients with basic literacy and some in regional languages.
Then a two-option question, plainly worded: how do you feel after the test, fine or uncomfortable. The clinician in a corridor observes distress. Here the patient reports it, and the reporting is structured rather than free text so it arrives at the doctor as data rather than as a note nobody reads.
SpO2, blood pressure and heart rate are captured after the walk, from a connected device or entered manually, because a household without a paired device should not be locked out of the programme.
This is how the empathy-versus-signal tension resolves, and it resolves structurally rather than through copy.
Distance covered and the change since last test — 400m, 50m more than last time. Never a clinical threshold, never a grade, never "below a norm." Progress-relative, always.
The full report — absolute distance, SpO2, blood pressure, heart rate, and the symptoms reported.
So the patient is not asked to self-assess a number they have no context for, and no clinically meaningful event depends on them raising the alarm. Calm for the patient, complete for the clinician, no trade between them. History is presented the same way — the patient's own distance over six months or a year, in their own trajectory rather than against a benchmark.
Different problem, same discipline. The stated objective was to increase the logging of medicines and other tasks by patients.
Patients were not logging their medication even when they had taken it and had been reminded to log it. Multiple medicines due at the same time produced a wall of separate cards. Marking each one done was laborious, and the volume of cards buried everything upcoming.
So the failure was not forgetting to take medication. It was the cost of reporting medication that had already been taken. The app was losing adherence data on patients who were adhering.
What I did about the gap in research access. I had no route to patients, so I used my father, who is elderly and takes multiple daily medications, to understand the mechanics: how doses get organised, where the confusion happens, what "did I already take that one" feels like. I want to be precise about what that is and is not. It gave me the physical and cognitive reality of an older person managing a regimen. It told me nothing about post-cardiac fear, clinical protocol, or how a heart patient relates to a recovery programme. It was a source for one specific question, not a substitute for patient research.
Patients think "before breakfast", not "09:00". So doses group into morning, afternoon, evening and night, with before-and-after-meal sub-groupings inside them. The schedule now matches how the patient already holds it in their head.
Patients take eight or nine medicines at once, and usually take all of them. Inverting it — one tap for "taken all", and a "skipped few" path that opens a picker only when something was missed — means the common case costs one action and the rare case costs two.
That inversion is the whole redesign. It also improves the clinical data rather than trading it away, because a skipped dose is now an explicit, itemised report instead of an absence that could mean anything.
I also proposed smart-speaker voice logging for patients who struggle with a phone. It was still in discussion when I left.
The vitals and assessment work carries the same tension as the walk test, resolved the same way. Weekly assessment shows adherence and vitals as counts in and out of range, charted over time. What varies is the accompanying line, which is written against thresholds rather than fixed.
Straightforward praise.
"You're on the right path."
A direct but non-punitive nudge about taking medicines on time.
The point of tiering the copy is that a patient having a bad health week is the patient most likely to abandon the programme, and the one whose data the care team most needs. Encouragement calibrated to performance keeps them reporting. Undifferentiated praise would make the whole thing meaningless, and undifferentiated warning would drive them away.
Streaks, milestones at seven, fourteen and thirty days, a badge system across medication, vitals, steps and sleep, and an onboarding flow to introduce it. I ran competitive research, built the progression logic, and designed the visuals and motion. It had not gone to development when I left.
I want to be straight about this one. It arrived as a solution from a stakeholder session rather than from a diagnosed problem, and gamifying adherence in a cardiac programme is genuinely contestable. Streaks reward continuity, and a patient who is unwell for a week loses the streak precisely when they are most fragile. There is a real risk of rewarding the appearance of adherence over adherence itself.
Uploading a prescription, connecting a tracker, starting with a health coach — so progress is available to a patient having a bad clinical week.
Previous streaks are preserved and displayed rather than erased, so a broken streak reads as history rather than failure.
Whether that is sufficient, I do not know, because it never reached patients. If I were doing it again I would want the clinical team's view on streak mechanics before designing them, in the same way I went to a cardiologist before designing the walk test.
I have no post-launch data. The engagement ended as the work went live. That pattern holds across my client work, and it is worth naming once rather than repeating: as an embedded designer placed with a client, I am typically scoped to delivery, and the measurement window opens after the engagement closes. It is a structural feature of how I have worked, not a lack of interest in results. Where I could chase numbers afterwards I have, and where I could not, I say so.
What I can evidence is the reasoning and the artefacts. What I cannot tell you is whether adherence moved.
I designed a clinical assessment for heart failure patients without ever watching one attempt it. The clinician session was the right compensating move and I would do it again, but it is compensation, not a substitute. One session observing a patient take this test at home would have taught me things no one in a meeting would think to mention.
The safety of this flow depends on a patient with basic literacy correctly understanding when to stop. That is a testable assumption and I never tested it.
When you move a supervised activity into an unsupervised setting, the interface is not the thing you are designing. You are redesigning who is responsible for the person — and you have to name the replacement explicitly.
Here it was a caretaker in the room, a screening gate, and a reporting path that does not depend on the patient. Everything else was detail.